前方注視カメラを活用した軌道の曲率推定

抄録

Active steering is one of the methods to improve the curve passing performance of railway vehicle. It improves steering performance by driving the actuator in accordance with the curvature. Curvature is a necessary parameter for active steering. This paper proposes a method for estimating curvature of track using a front view camera. The proposed system can be divided into three parts. First, semantic segmentation is performed to estimate the position of track from camera images. Second, the track position and radius are calculated by performing circular and straight-line fitting based on the estimated track position used semantic segmentation. Finally, the proposed system calculates the mean squared error of the track position estimated by semantic segmentation and the track position calculated by circular or straight-line fitting, and the proposed system is determined whether the track section is a straight or curved section. Furthermore, the proposed system is verified by using a 1/10 scale model experimental vehicle. Validation results show that the proposed system is capable of estimating curvature of track.